Drug spectrum and disproportionality signals for diabetes insipidus in FAERS: multi-method analysis, JADER cross-database assessment, and PubMed-based case-level evidence

Figures

• 

  FIGURE 1. Study design and analytical framework. Integrated workflow for identifying and prioritizing drug-associated diabetes insipidus (DI) pharmacovigilance signals using FAERS, JADER, and published case-level evidence. FAERS reports from 2004Q1 to 2025Q4 were used for the primary analysis, and JADER reports from 2004 to 2025 were used for cross-database assessment. DI cases were defined using predefined MedDRA Preferred Terms based on a MedDRA version 26.1 term audit, including “Diabetes insipidus” and “Nephrogenic diabetes insipidus.” Drug exposure was restricted to non-treatment primary suspect drugs after exclusion of drugs used for DI diagnosis or treatment, and drug names were harmonized to generic active ingredients. The main FAERS signal set focused on the 50 non-treatment primary suspect drugs with the highest DI-related reporting frequencies and was evaluated using crude ROR, adjusted ROR, EBGM, and IC. Sensitivity analyses included PS + SS analysis, PRR analysis, ROR-ranked prioritization, clinical-context exclusion using indication and co-reported adverse-event fields, and PT-level subtype assessment. The same FAERS frequency-ranked top 50 drugs were assessed in JADER using…

• 

  FIGURE 2. PRISMA flow diagram for published case-level evidence. Flow diagram showing the identification, screening, retrieval, eligibility assessment, and inclusion of published case-level evidence on suspected drug-associated diabetes insipidus (DI) or DI-like polyuria. Expanded database searches and supplemental drug-targeted searches identified 3,689 records, of which 787 duplicates were removed. After deduplication, 2,902 records were screened, and 1,180 records were marked for full-text review by automated screening. Eighty-one reports were sought for retrieval, 30 were retrieved and assessed for eligibility, and 29 publications comprising 32 individual cases were included in the systematic case-level review. The published-case synthesis was used to summarize extractable clinical features and diagnostic evidence, rather than to estimate incidence, comparative risk, or causality.

• 

  FIGURE 3. Demographic, geographic, and temporal distribution of diabetes insipidus-related reports in FAERS from 2004 to 2025. (A) Heat map showing the distribution of diabetes insipidus (DI)-related individual case safety reports by age group and sex. Numbers within each cell indicate the number of reports in the corresponding age–sex subgroup. (B) Distribution of DI-related reports across the five most frequently reported countries or regions. Numbers beside the bars indicate report counts. (C) Annual number of DI-related reports submitted to FAERS between 2004 and 2025.

• 

  FIGURE 4. ATC distribution of the frequency-ranked top 50 FAERS non-treatment drugs associated with diabetes insipidus. The donut chart shows the distribution of the 50 non-treatment primary suspect drugs with the highest number of diabetes insipidus (DI)-related reports in FAERS after exclusion of drugs used for DI diagnosis or treatment. Drugs were grouped according to first-level Anatomical Therapeutic Chemical (ATC) classification. Labels indicate the number and percentage of drugs in each therapeutic category among the frequency-ranked top 50. Nervous system agents accounted for the largest proportion, followed by antineoplastic and immunomodulating agents, systemic hormonal agents, anti-infectives for systemic use, cardiovascular system drugs, and other categories.

• 

  FIGURE 5. Multi-method disproportionality assessment of the frequency-ranked top 50 FAERS non-treatment primary suspect drugs associated with diabetes insipidus. The four-panel forest-bubble plot compares crude reporting odds ratio (ROR), covariate-adjusted ROR, empirical Bayes geometric mean (EBGM), and information component transformed to the reporting ratio scale [IC (2IC)] for the 50 non-treatment primary suspect drugs with the highest number of diabetes insipidus (DI)-related reports in FAERS. Drugs used for DI diagnosis or treatment were excluded before signal calculation and top-50 selection. The x-axis is displayed on a logarithmic scale, and the vertical dotted line indicates the null value of 1.0. Points indicate the corresponding disproportionality estimate, horizontal lines indicate 95% confidence intervals for ROR estimates or uncertainty intervals for shrinkage-based estimates, point size represents the number of FAERS DI-related reports, and point color represents first-level Anatomical Therapeutic Chemical (ATC) class. Point shape indicates whether the predefined signal criterion for each method was met.

• 

  FIGURE 6. FAERS top 50 non-treatment drugs by diabetes insipidus-related report count, disproportionality estimates, and frequency-based selection context. (A) Number of DI-related reports for the top 50 non-treatment primary suspect drugs in FAERS, ranked by report count and colored according to the first-level ATC class. (B) Crude RORs and corresponding 95% confidence intervals for the same 50 drugs, displayed on a logarithmic scale. The dashed vertical line indicates the null value of ROR = 1. (C) Contextual distribution of report count and disproportionality among eligible non-treatment drugs in FAERS. Grey points represent all eligible drugs, whereas colored points highlight the top 50 drugs using the same ATC classification as in (A,B).

• 

  FIGURE 7. Demographic and temporal distribution of diabetes insipidus-related reports in the JADER database from 2004 to 2025. (A) Heat map showing the distribution of diabetes insipidus (DI)-related individual case safety reports by age group and sex in the Japanese Adverse Drug Event Report (JADER) database. Numbers within each cell indicate the number of reports in the corresponding age–sex subgroup. (B) Annual number of DI-related reports recorded in JADER between 2004 and 2025.

• 

  FIGURE 8. Cross-database assessment of FAERS-derived non-treatment diabetes insipidus signals in JADER. The four-panel forest-bubble plot shows crude ROR, covariate-adjusted ROR, EBGM, and IC transformed to the reporting ratio scale [IC (2IC)] calculated in JADER for drugs included in the FAERS frequency-ranked top-50 non-treatment primary suspect drug set. The x-axis is displayed on a logarithmic scale, and the vertical dashed line indicates the null value of 1.0. Points indicate the corresponding JADER disproportionality estimate, horizontal lines indicate 95% confidence intervals for ROR estimates or uncertainty intervals for shrinkage-based estimates, point size represents the number of JADER DI-related reports, and point color represents the first-level Anatomical Therapeutic Chemical (ATC) class. Point shape indicates whether the predefined signal criterion for each method was met. This analysis was used to assess cross-database signal reproducibility rather than to confirm causality.

• 

  FIGURE 9. Cross-database comparison of diabetes insipidus-related adjusted disproportionality signals between FAERS and JADER. Scatter plot comparing adjusted reporting odds ratio (adjusted ROR) estimates between FAERS and JADER for 48 drugs with finite estimable adjusted ROR values in both databases among the 50 frequency-ranked non-treatment primary suspect drugs. Both axes are displayed on logarithmic scales. The x-axis shows FAERS adjusted ROR estimates, and the y-axis shows JADER adjusted ROR estimates. The dashed diagonal line indicates equal adjusted ROR estimates between the two databases, and the dotted reference lines indicate the null threshold of adjusted ROR = 1.0. Points are colored according to the first-level Anatomical Therapeutic Chemical (ATC) class. Point shapes indicate JADER cross-database support status; robust support was defined as positivity in at least three of the four JADER methods, namely, crude ROR, adjusted ROR, EBGM, and IC. The smoothed curve and shaded band show the overall trend and uncertainty. Spearman correlation was used to assess concordance in adjusted ROR magnitude across databases. Cross-database comparison reflects reproducibility of…

• 

  FIGURE 10. Cross-database comparison of diabetes insipidus-related disproportionality estimates using four analytic methods. Forest plot comparing FAERS and JADER disproportionality estimates for FAERS frequency-ranked top 50 non-treatment primary suspect drugs with at least three diabetes insipidus (DI)-related reports in JADER. The four panels show crude ROR, covariate-adjusted ROR, EBGM, and IC transformed to the reporting ratio scale [IC (2IC)]. The x-axis is displayed on a logarithmic scale, and the dashed vertical line indicates the null value of 1.0. Points represent database-specific estimates, with within-point labels indicating the database source: F for FAERS and J for JADER. Horizontal lines represent 95% confidence intervals for ROR-based estimates or uncertainty intervals for EBGM and IC-based estimates. Solid interval lines indicate positive signals, whereas dashed interval lines indicate non-positive signals according to the corresponding method-specific criterion. Colors indicate the first-level Anatomical Therapeutic Chemical (ATC) class. This comparison summarizes cross-database consistency across crude, adjusted, and shrinkage-based disproportionality methods and should…

• 

  FIGURE 11. PubMed case-level evidence map for suspected drug-associated diabetes insipidus. Drug-level summary of published case evidence for suspected drug-associated diabetes insipidus (DI) or DI-like polyuria. Rows represent suspected drugs, and columns summarize drug class, exposure-attributed published case count, main reported DI phenotype, diagnostic features, dechallenge, rechallenge, and causality assessment. Bubble size in the “Cases” column reflects the number of exposure-attributed published cases for each drug. Diagnostic features are shown as the number of cases reporting each feature divided by the total number of published cases for that drug; filled circles indicate more frequently reported features, and open circles indicate absent or less frequently reported features. Abbreviations: UO, polyuria or urine output; Na, serum sodium; Uosm, urine osmolality or specific gravity; Sosm, serum osmolality; DDAVP, desmopressin or vasopressin response. The figure summarizes published case-level clinical evidence and should not be interpreted as estimating incidence, comparative risk, or causality.